What synthetic aperture radar actually sees
Every other Earth observation instrument waits for the world to cooperate. Radar brings its own light, and that one difference decides what it is good for and what it is useless at.
· 9 minute read · Caelus
Start with the failure, because the failure is what makes the case. A cyclone crosses an island at two in the morning. By the time anyone can fly, the weather system is still sitting on top of the place, and it will sit there for three more days. An optical satellite passes overhead on schedule, takes a beautiful photograph of the top of a cloud, and moves on. It does that again the next day. And the next.
This is not a rare edge case. It is the normal condition of most disasters, because the weather that causes the disaster is the same weather that hides it. Floods come with cloud. Fires come with smoke. Storms arrive at night. The instruments that carry the most public trust, the ones that produce the images everybody recognises, are blind in exactly the hours their output would matter most.
Optical imaging is photography. It needs the sun, and it needs a clear line of sight. Radar needs neither, because it brings its own illumination.
The trick is that radar is not a camera
A camera is a passive instrument. It collects light that something else produced, which for an Earth observation satellite means sunlight that bounced off the ground. Take the sun away and you have nothing. Put a cloud in the path and you have a picture of the cloud.
Synthetic aperture radar is an active instrument. The satellite transmits a microwave pulse at the ground and listens for what comes back. It is measuring its own signal, so the sun is irrelevant: SAR works at midnight exactly as well as at noon. And the wavelengths involved are centimetres rather than hundreds of nanometres, which is far larger than a cloud droplet. The pulse goes straight through weather that optical light cannot penetrate at all.
That is the entire superpower, and it is worth being precise about why it matters. It is not that radar is sharper, or newer, or more sophisticated. It is that radar is available. A twelve-day revisit that actually returns data every twelve days beats a five-day revisit that returns cloud for six weeks of monsoon.
The word synthetic is doing real work
Radar resolution depends on antenna size. To resolve detail from orbit at these wavelengths you would need an antenna kilometres long, which is not a thing anybody is launching.
So SAR cheats, brilliantly. The satellite is moving at roughly seven kilometres a second. It fires pulses continuously as it travels, and each pulse comes back from the same patch of ground at a slightly different angle. Process those returns together, accounting for the shift in the satellite's position and the Doppler shift in the returns themselves, and you have synthesised a virtual antenna as long as the distance the satellite flew while that patch was in view. A ten metre antenna behaves like a ten kilometre one. That is the synthetic aperture, and it is the reason a spacecraft the size of a van can resolve a building.
What it is measuring is texture, not colour
An optical image records reflectance across visible and near-infrared bands, which is why it looks like the world. A SAR image records backscatter: how much of the transmitted pulse came back toward the satellite. And what governs backscatter is geometry and material, not colour.
- Smooth flat surfaces reflect the pulse away from the satellite like a mirror angled wrong. Calm open water comes back almost black.
- Rough surfaces scatter in all directions, including back. Bare soil and vegetation come back mid-grey.
- Vertical structures meeting the ground form a corner that returns the pulse almost perfectly. A building against a street comes back brilliantly bright.
- Wet material returns more than dry material, because water changes how the surface responds to microwaves.
Read that list again with a flood in mind and the appeal becomes obvious. Standing water is the darkest thing in the scene, and buildings are the brightest. Flood extent in a SAR image is not something you infer delicately from subtle spectral differences. It is the part of the picture that went black.
A flood in an optical image is a judgement call about colour. A flood in a radar image is a hole.
So what does everything else do better
Plenty, and pretending otherwise is how people end up using the wrong instrument for years. Here is the honest comparison.
| Instrument | Strength | The catch |
|---|---|---|
| SAR (radar) | Sees through cloud, smoke and darkness. Detects structural change, surface water, and ground motion down to millimetres. | Hard to interpret. Arrives full of speckle. Terrain distorts geometry in mountains. Says nothing about colour, species, or health. |
| Optical and multispectral | Looks like the world, so anyone can read it. Spectral bands resolve vegetation health, water quality, crop type, burn severity. | Blind through cloud, smoke and night. In a monsoon region that can mean months with no usable scene. |
| Thermal infrared | Measures emitted heat. Finds fires, gas flares, urban heat, and irrigation stress directly. | Coarse resolution. Cloud blocks it too. Heat is an ambiguous signal on its own. |
| Lidar | Genuine three-dimensional structure. Canopy height and bare-earth elevation under vegetation, at a precision nothing else touches. | Almost always airborne, so coverage is a flight campaign, not a global archive. Expensive, and rarely repeated often enough to show change. |
The real answer is that you use more than one
The instinct to pick a favourite instrument is the instinct to have a hammer. In practice the interesting findings almost always come from disagreement between two sensors that were built to measure different things.
The Thar Desert is the cleanest example we have run into. Vegetation indices from optical imagery say western Rajasthan is greening, and they are right. Gravimetry from GRACE says the groundwater under it is disappearing, and that is right too. Read either alone and you reach a confident, reportable, completely wrong conclusion. Read them together and you have something specific: irrigated cultivation expanding into arid land, drawing on a store nobody is refilling. That is a timing problem, and it has a deadline.
Caelus builds around SAR because of availability, not because radar is the best sensor in the abstract. The communities this work is for are disproportionately in places with heavy cloud, active monsoons, smoke seasons, or night-time emergencies. The instrument that keeps returning data during the event is worth more than the instrument that produces a better picture afterward.
Where this came from.
Credited in plain text, never as logos. Use of open data is not a relationship, and a logo would imply one.
- Sentinel-1 mission specification
- European Space Agency. C-band SAR, twelve-day repeat cycle per satellite.
- Landsat and Sentinel-2 programmes
- USGS and ESA. The optical archives that reach back furthest.
- GRACE and GRACE-FO
- NASA and GFZ. Terrestrial water storage anomaly from gravimetry.